A straightforward line search approach on the expected empirical loss for stochastic deep learning problems.
Maximus MutschlerAndreas ZellPublished in: CoRR (2020)
Keyphrases
- learning problems
- line search
- learning tasks
- machine learning
- learning algorithm
- supervised learning
- machine learning algorithms
- conjugate gradient
- step size
- kernel methods
- quadratic programming
- convergence rate
- objective function
- risk minimization
- global convergence
- primal dual
- semi supervised learning
- reinforcement learning
- reproducing kernel hilbert space
- kernel machines
- neural network
- loss function
- learning theory
- multi task